| Metric | naive | lstm | cnn | xgboost |
|---|---|---|---|---|
| diff_of_means | 15.057 | 18.676 | 20.325 | 22.959 |
| ratio_of_sd | 0.831 | 0.925 | 0.888 | 0.841 |
| amplitude_ratio_of_means | 0.588 | 0.701 | 0.646 | 0.581 |
| maximum_error | 0.329 | 0.339 | 0.308 | 0.302 |
| ks_mean_on_coarse_res_with_extremes | 0.495 | 0.154 | 0.228 | 0.328 |
| rainy_hours_ratio_of_means | 0.973 | 1.259 | 1.286 | 1.208 |
| qqplot_mae | 0.032 | 0.039 | 0.042 | 0.042 |
| acf_mae | 0.128 | 0.062 | 0.090 | 0.106 |
| extremogram_mae | 0.109 | 0.030 | 0.052 | 0.063 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | naive | lstm | cnn | xgboost |
|---|---|---|---|---|
| diff_of_means | 15.975 | 19.366 | 20.981 | 24.201 |
| ratio_of_sd | 0.850 | 0.948 | 0.902 | 0.852 |
| amplitude_ratio_of_means | 0.589 | 0.703 | 0.651 | 0.581 |
| maximum_error | 0.342 | 0.354 | 0.323 | 0.320 |
| ks_mean_on_coarse_res_with_extremes | 0.510 | 0.194 | 0.178 | 0.343 |
| rainy_hours_ratio_of_means | 1.002 | 1.294 | 1.315 | 1.244 |
| qqplot_mae | 0.032 | 0.041 | 0.043 | 0.044 |
| acf_mae | 0.131 | 0.060 | 0.094 | 0.104 |
| extremogram_mae | 0.106 | 0.037 | 0.054 | 0.064 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | naive | lstm | cnn | xgboost |
|---|---|---|---|---|
| diff_of_means | 11.555 | 15.918 | 16.316 | 19.591 |
| ratio_of_sd | 0.865 | 0.952 | 0.923 | 0.876 |
| amplitude_ratio_of_means | 0.603 | 0.737 | 0.693 | 0.609 |
| maximum_error | 0.335 | 0.346 | 0.322 | 0.309 |
| ks_mean_on_coarse_res_with_extremes | 0.485 | 0.159 | 0.165 | 0.328 |
| rainy_hours_ratio_of_means | 0.968 | 1.256 | 1.270 | 1.205 |
| qqplot_mae | 0.027 | 0.033 | 0.036 | 0.036 |
| acf_mae | 0.134 | 0.060 | 0.087 | 0.105 |
| extremogram_mae | 0.099 | 0.024 | 0.042 | 0.057 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | xgboost | cnn | lstm | naive |
|---|---|---|---|---|
| diff_of_means | -14.565 | -15.289 | -16.518 | -17.812 |
| ratio_of_sd | 0.938 | 0.941 | 0.992 | 0.864 |
| amplitude_ratio_of_means | 0.781 | 0.860 | 0.908 | 0.663 |
| maximum_error | 0.284 | 0.264 | 0.295 | 0.335 |
| ks_mean_on_coarse_res_with_extremes | 0.202 | 0.119 | 0.076 | 0.531 |
| rainy_hours_ratio_of_means | 0.805 | 0.848 | 0.836 | 0.704 |
| qqplot_mae | 0.039 | 0.039 | 0.035 | 0.057 |
| acf_mae | 0.130 | 0.115 | 0.086 | 0.173 |
| extremogram_mae | 0.045 | 0.024 | 0.014 | 0.080 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | naive | lstm | cnn | xgboost |
|---|---|---|---|---|
| diff_of_means | 7.742 | 9.041 | 13.846 | 15.356 |
| ratio_of_sd | 0.853 | 0.959 | 0.898 | 0.886 |
| amplitude_ratio_of_means | 0.609 | 0.775 | 0.726 | 0.616 |
| maximum_error | 0.339 | 0.403 | 0.353 | 0.290 |
| ks_mean_on_coarse_res_with_extremes | 0.496 | 0.153 | 0.160 | 0.293 |
| rainy_hours_ratio_of_means | 0.924 | 1.133 | 1.185 | 1.140 |
| qqplot_mae | 0.025 | 0.021 | 0.030 | 0.028 |
| acf_mae | 0.144 | 0.071 | 0.086 | 0.115 |
| extremogram_mae | 0.086 | 0.020 | 0.028 | 0.058 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | naive | lstm | cnn | xgboost |
|---|---|---|---|---|
| diff_of_means | 0.673 | 2.575 | 6.338 | 8.413 |
| ratio_of_sd | 0.899 | 1.004 | 0.950 | 0.938 |
| amplitude_ratio_of_means | 0.632 | 0.819 | 0.782 | 0.659 |
| maximum_error | 0.321 | 0.361 | 0.334 | 0.278 |
| ks_mean_on_coarse_res_with_extremes | 0.565 | 0.167 | 0.161 | 0.324 |
| rainy_hours_ratio_of_means | 0.893 | 1.093 | 1.136 | 1.094 |
| qqplot_mae | 0.020 | 0.013 | 0.018 | 0.015 |
| acf_mae | 0.158 | 0.079 | 0.092 | 0.122 |
| extremogram_mae | 0.107 | 0.024 | 0.037 | 0.067 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97
| Metric | naive | lstm | cnn | xgboost |
|---|---|---|---|---|
| diff_of_means | 7.345 | 9.426 | 13.505 | 16.281 |
| ratio_of_sd | 0.901 | 0.977 | 0.942 | 0.898 |
| amplitude_ratio_of_means | 0.612 | 0.742 | 0.708 | 0.588 |
| maximum_error | 0.333 | 0.376 | 0.352 | 0.294 |
| ks_mean_on_coarse_res_with_extremes | 0.507 | 0.223 | 0.202 | 0.352 |
| rainy_hours_ratio_of_means | 0.950 | 1.154 | 1.218 | 1.185 |
| qqplot_mae | 0.020 | 0.022 | 0.031 | 0.030 |
| acf_mae | 0.140 | 0.086 | 0.090 | 0.125 |
| extremogram_mae | 0.094 | 0.035 | 0.040 | 0.059 |
Important: Right now we are only estimating the upper tail
extremogram. Currently we didn’t find a way to estimate the two tales at
the same time. We are using quant = .97